AssemblyAI

Overview

AssemblyAI integration in Mosaico provides automated speech-to-text transcription capabilities for audio assets. This integration enables accurate transcription with word-level timing, which is essential for subtitle generation and content synchronization.

Requirements

• AssemblyAI Python package (pip install assemblyai)
• Valid AssemblyAI API key
• Audio in a supported format (MP3, WAV, etc.)

Usage

from mosaico.audio_transcribers import AssemblyAIAudioTranscriber
from mosaico.assets.audio import AudioAsset
from mosaico.video.project import VideoProject

# Initialize transcriber
transcriber = AssemblyAIAudioTranscriber(
    api_key="your_api_key",
    model="best",  # or "nano" for faster processing
    language="en"  # optional language specification
)

# Create audio asset
audio = AudioAsset.from_path("narration.mp3")

# Transcribe audio
transcription = transcriber.transcribe(audio)

# Access transcription results
for word in transcription.words:
    print(f"{word.text}: {word.start_time} - {word.end_time}")

Configuration Options

The AssemblyAIAudioTranscriber supports several configuration options:

• api_key: Your AssemblyAI API key (required)
• model: Transcription model to use (best or nano)
• language: Optional language specification
• custom_spelling: Dictionary of custom spelling corrections

Features

Language Detection

# Automatic language detection
transcriber = AssemblyAIAudioTranscriber(
    api_key="your_api_key",
    language=None  # Enables automatic detection
)

Custom Spelling

# Add custom spelling corrections
transcriber = AssemblyAIAudioTranscriber(
    api_key="your_api_key",
    custom_spelling={
        "mosaico": "Mosaico",
        "ai": ["AI", "A.I."]
    }
)

Integration with Video Projects

The transcription can be used in multiple ways with video projects:

# Create video project
project = VideoProject()

# Add audio asset
project.add_assets(audio_asset)

# Add captions from transcriber
project = project.add_captions_from_transcriber(
    transcriber,
    max_duration=5,  # Maximum subtitle duration
    params=TextAssetParams(
        font_size=36,
        font_color="white"
    ),
    overwrite=False  # Don't overwrite existing captions
)

# Or manually add captions from transcription
project = project.add_captions(
    transcription,
    max_duration=5,
    params=TextAssetParams(
        font_size=36,
        font_color="white"
    ),
    scene_index=0,  # Add to specific scene
    overwrite=True  # Replace existing captions
)